Re: [PATCH net-next v2 1/2] net: use two lockdep classes for the netdev instance lock

[Paolo Abeni <pabeni@redhat.com>]

On 9/26/26 06:18, Jakub Kicinski wrote:
> diff --git a/net/core/dev.c b/net/core/dev.c
> index f660fccfc0db..a8eb382f40ca 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -568,6 +568,41 @@ static inline void netdev_set_addr_lockdep_class(struct net_device *dev)
>   }
>   #endif
>   
> +#ifdef CONFIG_PROVE_LOCKING
> +static int netdev_lock_cmp_fn(const struct lockdep_map *a,
> +			      const struct lockdep_map *b)
> +{
> +	if (a == b)
> +		return 0;
> +
> +	/* @a and @b must be of same class - both virtual or physical.
> +	 * cmp_fn won't be called for devices of different classes.
> +	 *
> +	 * For the same class only allow nesting under the protection
> +	 * of rtnl_lock. Note that we can't use lockdep_rtnl_is_held()
> +	 * here, it always answers UNKNOWN from within lockdep.
> +	 */
> +	return rtnl_is_locked() ? -1 : 1;
> +}
> +
> +/* A virtual device can be locked before the physical device it leases
> + * queues from, see netdev_nl_queue_create_doit(). Keep the two kinds
> + * in separate classes so the dependency graph enforces the order;
> + * netdev_lock_cmp_fn() then only has to rule on same-class nesting.
> + */
> +void netdev_set_instance_lock_class(struct net_device *dev)
> +{
> +	static struct lock_class_key netdev_virt_instance_lock_key;
> +
> +	if (dev->dev.parent)
> +		return;
> +
> +	lockdep_set_class(&dev->lock, &netdev_virt_instance_lock_key);
> +	lock_set_cmp_fn(&dev->lock, netdev_lock_cmp_fn, NULL);
> +}

Sashiko noted that mlx5 IB should enter !dev->dev.parent code path and
could possibly need a follow-up.

/P